Skip to main content

Method Library

The Method Library stores reusable data processing methods for DFS Pro. Use it when a fusion task or governance workflow needs repeatable logic, such as preprocessing, feature extraction, merge logic, labeling, or math operations.

Open:

Data Integration > Method Library

Method categories​

CategoryTypical use
PreprocessingClean values, normalize units, reshape rows, prepare fields.
Feature ExtractionCreate derived features for analytics or predictive maintenance.
MergeCombine records from several datasets.
LabelAdd labels, classes, status flags, or review tags.
MathApply calculations, thresholds, scoring, or aggregation logic.

Built-in and custom methods​

Built-in methods are provided by the platform. Custom methods are created by the tenant team for project-specific logic.

Use custom methods when:

  • source data uses project-specific fields;
  • an output must follow a customer naming convention;
  • a fusion task needs repeatable matching behavior;
  • predictive maintenance features need documented preparation logic;
  • a reviewer needs to test logic before using it in a workflow.

Create a custom method​

  1. Open Method Library.
  2. Select Add Method.
  3. Enter method name.
  4. Choose category.
  5. Add description.
  6. Add pseudo code that explains the logic.
  7. Add sample input JSON.
  8. Add expected output JSON when the result should be compared.
  9. Save the method.

Use pseudo code that a reviewer can understand. It should explain the operational logic, not only the implementation detail.

Test a method​

Test a method before publishing or using it in a fusion task.

  1. Open the method.
  2. Select Test.
  3. Enter sample input JSON.
  4. Enter expected output if available.
  5. Run test.
  6. Review output, duration, rows in, rows out, and match result.

Good sample input includes:

  • normal records;
  • missing fields;
  • boundary values;
  • duplicate keys;
  • values that should be rejected or flagged;
  • source-specific examples from a real workflow.

Publish a method​

Publish only after sample tests are acceptable.

Before publishing, confirm:

  • method name is clear;
  • category is correct;
  • description explains when to use it;
  • pseudo code is understandable;
  • sample input is representative;
  • expected output is correct;
  • reviewer accepts the behavior.

Published methods can be used by downstream workflows such as fusion tasks.

Create a new version​

Create a new version when method logic changes.

Use a new version when:

  • matching keys change;
  • feature calculation changes;
  • source priorities change;
  • tolerance or threshold changes;
  • output schema changes;
  • downstream workflow owners need repeatability.

Before deprecating an older version, check method usage and affected fusion tasks.

Monitor usage and performance​

Method Library can show usage and performance information.

Review:

  • fusion task usage count;
  • harness or test runs;
  • success rate;
  • p50, p95, and p99 latency when available;
  • recent failures.

Use this information before editing, versioning, or deprecating a method.

Example: predictive maintenance feature method​

A feature extraction method might:

  1. read equipment ID, timestamp, vibration, temperature, and current;
  2. normalize units;
  3. calculate rolling average or rate of change;
  4. flag missing intervals;
  5. return a feature table for predictive maintenance review.

The method should be tested with sample input that includes normal data, missing intervals, and out-of-range values.

PageUse
DFS Pro DatasetsPrepare input datasets for methods and fusion tasks.
Fusion TasksUse published methods in multi-source fusion.
Review QueueReview uncertain outputs produced by fusion workflows.